Splitting Dataframe Into Two And Using Tilde ~ As Variable

I wanna do 2 similar operations with Pandas in Python 3. One with tilde and another without tilde. 1 - df = df[~(df.teste.isin(['Place']))] 2 - df = df[(df.teste.isin(['Place']))]

Solution 1:

For your limited use case, there is limited benefit in what you are requesting.

GroupBy

Your real problem, however, is the number of variables you are having to create. You can halve them via GroupBy and a calculated grouper:

df = pd.DataFrame({'teste': ['Place', 'Null', 'Something', 'Place'],
                   'value': [1, 2, 3, 4]})

dfs = dict(tuple(df.groupby(df['teste'] == 'Place')))

{False:        teste  value
        1Null22Something3,

 True:         teste  value
            0Place13Place4}

Then access your dataframes via dfs[0] and dfs[1], since False == 0 and True == 1. There is a benefit with this last example. You now remove the need to create new variables unnecessarily. Your dataframes are organized since they exist in the same dictionary.

Function dispatching

Your precise requirement can be met via the operator module and an identity function:

from operator import invert

tilde = [invert, lambda x: x]

mask = df.teste == 'Place'  # don't repeat mask calculations unnecessarily

df1 = df[tilde[0](mask)]
df2 = df[tilde[1](mask)]

Sequence unpacking

If your intention is to use one line, use sequence unpacking:

df1, df2 = (df[func(mask)] forfuncintilde)

Note you can replicate the GroupBy result via:

dfs = dict(enumerate(df[func(mask)] forfuncintilde)

But this is verbose and convoluted. Stick with the GroupBy solution.

Solution 2:

You could possibly condense your code a little by defining your tests and then iterating over those. Let me illustrate:

tests = ["Place", "Foo", "Bar"]

for t in tests:
    # not sure what you are doing exactly, just copied it
    1 - df = df[~(df.teste.isin([t]))] 
    2 - df = df[(df.teste.isin([t]))]

That way, you only have two linesdoing the actual work, and simply adding another test to the list saves you duplicating code. No idea if this is what you want, though.